EP0704810A1 - Procédé et dispositif pour classer des informations de documents - Google Patents

Procédé et dispositif pour classer des informations de documents Download PDF

Info

Publication number
EP0704810A1
EP0704810A1 EP95115253A EP95115253A EP0704810A1 EP 0704810 A1 EP0704810 A1 EP 0704810A1 EP 95115253 A EP95115253 A EP 95115253A EP 95115253 A EP95115253 A EP 95115253A EP 0704810 A1 EP0704810 A1 EP 0704810A1
Authority
EP
European Patent Office
Prior art keywords
key word
document information
folder
document
folders
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
EP95115253A
Other languages
German (de)
English (en)
Other versions
EP0704810B1 (fr
Inventor
Takako Morita, (Nee Sakai)
Junichi Higashino
Yoshiki Matsuda
Tetsuya Hashimoto
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hitachi Ltd
Original Assignee
Hitachi Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hitachi Ltd filed Critical Hitachi Ltd
Publication of EP0704810A1 publication Critical patent/EP0704810A1/fr
Application granted granted Critical
Publication of EP0704810B1 publication Critical patent/EP0704810B1/fr
Anticipated expiration legal-status Critical
Expired - Lifetime legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification
    • G06F16/355—Class or cluster creation or modification
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
    • Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y10S707/00—Data processing: database and file management or data structures
    • Y10S707/99931—Database or file accessing
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
    • Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y10S707/00—Data processing: database and file management or data structures
    • Y10S707/99931—Database or file accessing
    • Y10S707/99932—Access augmentation or optimizing
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
    • Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y10S707/00—Data processing: database and file management or data structures
    • Y10S707/99931—Database or file accessing
    • Y10S707/99933—Query processing, i.e. searching
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y10—TECHNICAL SUBJECTS COVERED BY FORMER USPC
    • Y10S—TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y10S707/00—Data processing: database and file management or data structures
    • Y10S707/99931—Database or file accessing
    • Y10S707/99933—Query processing, i.e. searching
    • Y10S707/99934—Query formulation, input preparation, or translation

Definitions

  • the present invention relates to a method and an apparatus for classifying document files stored in a computer hierarchically. Particularly, the present invention is applied to an information classifying system and a document filing system.
  • Arrangement of a computer network is properly preparing environment in which collection of information is performed by utilizing an information retrieval system and an electronic mail. New information arrives successively through the computer network and topics or subjects are changed momently. Availability of the latest information is recognized sufficiently, while the latest information is not effectively utilized under the present conditions due to the following problems:
  • the wide-area information servers described in "Information Society in the 21st Century", Nikkei Byte, Nov. 1991, pp. 320-331 solves the problems in the above items (1) and (2) by the relevance feedback.
  • retrieval is performed on the basis of retrieval conditions designated by a user and when the user selects desired information from the retrieved results, the selected information is fed back to the retrieval conditions to improve the retrieval conditions.
  • learning is performed to polish the retrieval conditions.
  • a document information retrieval apparatus described in JP-A-5-28198 retrieves a document by using retrieval information such as classification given to each of documents and words appearing in the documents as key words.
  • retrieval data such as a field and a word which the user wants to retrieve
  • contents are presumed to prepare a retrieval expression automatically and perform retrieval to thereby solve the problem in the above item (1).
  • an optimum logical expression is selected from previously registered retrieval logical expressions. Further, similarities of the retrieved documents to the retrieval data are calculated and the documents are arranged in accordance with the calculated similarities to thereby cope with the problem in the item (3).
  • the system improves the retrieval condition by utilizing not only the retrieval condition designated by the user but also actual information satisfying a desire of the user by means of the relevance feedback method, so that the problems in the items (1) and (2) are solved.
  • the "Broad Catch System of Information” adopts the method in which a key word is registered without inputting a retrieval expression. However, since it is difficult for the user to register a key word which reflects the user's interest exactly, the problem in the item (2) is not solved. Further, it is premised that the automatic classification process of documents is started successively and accordingly the classification of documents is dependent on order of arrival of documents. Contents of information change momently and accordingly when the classification continues to be dependent on a classification system or a plurality of folders interrelated by a hierarchical relationship in a certain past time, an effective classification result cannot be produced and it becomes difficult to solve the problem in the item (3).
  • the classification process of documents has also the same problem in the work for previously establishing a classification system suggesting key words for classification. That is, if the classification continues to be dependent on one classification system, there is a problem that it cannot cope with variation of contents and an improper classification result is obtained.
  • retrieved documents are adapted to be arranged in order of availability thereof so that the documents can be reviewed in order of documents similar to a retrieval expression.
  • the user is required to confirm contents of the documents and accordingly the problem in the item (3) cannot be solved.
  • JP-A-5-324726 discloses measures for solving these problems.
  • attention is paid to a certain key word and a document is classified in accordance with whether the key word is described in the document or not.
  • the documents thus classified are subjected to the same process to thereby classify the documents hierarchically.
  • the present invention comprises storing a plurality of documents and a key word group in each document of the documents previously, clustering documents including a key word into a single key word folder or a folder stored with retrieval conditions for one word to be retrieved for each stored key word, comparing documents in the single key word folder with each other to combine the single key word folders with each other to prepare a related key word folder or a folder stored with retrieved conditions for designating a plurality of related word to be retrieved, and classifying the plurality of previously stored documents into the related key word folders.
  • the plurality of stored documents, the key word group including key words given to each document manually and words automatically extracted from each document, and the importance degree calculated from individual key words in the key word group are managed and the key word group and the importance degree are utilized for the classification process.
  • the document group is stored in the single key word folder by means of the single key word classification process of documents relating to each key word.
  • Documents having a plurality of key words are classified into a plurality of single key word folders in duplicate.
  • the integration is made on the basis of the related key word classification process and the key word folders are stored in the related key word folder.
  • Detailed classification judgment is performed for the inside of the single key word folder or the related key word folder.
  • the single key word process and the related key word classification process are utilized to hierarchically classify the inside of the folders.
  • the detailed classification is recurrently repeated until stop of classification is judged.
  • the detailed classification is impossible, the relevance between documents is judged and the document having the low relevance is regarded as noise and is separated.
  • a name representative of the document group stored in the folder is given to each related key word folder and a folder name is given thereto.
  • each document can be stored in one or more classification exactly without dependence on the existing classification system and the hierarchical classification system can be prepared.
  • the name representative of the classified result group can be given. Consequently, the user can easily detect a desired document from a large number of documents.
  • An object to be processed by a document classification apparatus of the first embodiment is text files in a computer and the text files are supposed to be documents.
  • a plurality of key words representative of contents of each document are given to each of documents and the key words are referred to as a key word group.
  • Fig. 1 schematically illustrates an example of the document classification apparatus of the first embodiment.
  • the document classification apparatus 100 comprises a data management unit 101, a word detection unit 102, a document classification unit 103, and a classified result output unit 104 and further includes a word dictionary 105 for storing common terms, a thesaurus 106 for storing the higher and lower rank relation and synonymous information among terms, a document data base (DB) 107 for storing documents, a key word data base (DB) 108 for storing key word groups of documents, a CRT 109 constituting an output device, a key board 110 and a mouse 111 constituting input devices.
  • DB document data base
  • DB key word data base
  • the data management unit 101 manages the document data base 107 and the key word data base 108 and performs input/output of documents and key word groups.
  • Key words given previously by hand (referred to as manually given key words) can be stored in the key word data base 108.
  • the manually given key words are not required to be stored in the data base, while in the first embodiment the case where the manually given key words are stored is described by way of example.
  • the word detection unit 102 receives a document from the data management unit 101 and performs the morphemic analysis by utilizing the word dictionary 105 to extract words from the document.
  • the word detection unit may utilize a method of dividing a document at each change point in types of characters to extract words, a method of examining the appearance frequency of a chunk of n characters in a document to extract the chunk of n characters having a meaning and the like.
  • the extracted words are developed to synonyms and are converted into higher-rank words with reference to the thesaurus to produce synonym groups.
  • the extracted words and the synonyms are combined to produce key words.
  • the key words are delivered to the data management unit 101 and are added to the key word groups of documents in the key word data base 108.
  • the word detection unit 102 calculates the degree of importance of each key word.
  • the degree of importance is calculated on the basis of (1) the frequency of appearance of each word in regard to the total number of the appearance frequencies of the whole words and (2) an appearance position of each word in the title, the headline or caption, the paragraph or the like of the document. That is, the possibility that the title represents contents of a document is high and accordingly the degree of importance of words appearing in the title is increased.
  • the document classification unit 103 receives the key word groups of each document from the data management unit 101 and prepares a hierarchical classification system, so that the results thereof are delivered to the classified result output unit 104. Detail of the classification process will be described later.
  • the classified result output unit 104 receives the hierarchical classification system from the document classification unit 103 and displays an interface picture in the CRT 109.
  • Fig. 2 shows a definite example of the key words.
  • the word detection unit 102 of Fig. 1 is described in detail.
  • a word group 202 is first extracted from a document 201 with reference to the word dictionary 105.
  • the word group 202 is developed to synonyms and higher-rank words with reference to the thesaurus 106.
  • a synonym group 203 illustrates an example where a term of "blended” is developed to a synonym of "mixed". The development to the synonym is made in order to unify expression or description of key words.
  • key words 204 are prepared from the word group 202 and the synonym group 203 and are stored in a key word group 205 of the document 201 stored in the key word data base 108.
  • the key word group 205 includes manually given key words and the key words 204 added thereto.
  • An area in which the document is stored as a result of classification is named a folder and a plurality of documents stored in the folder is named a document group. Further, "preparing the folder” means ensuring an area for storing the classified document group.
  • a single key word classification process 301 for clustering documents having key words included in the key word group for each key word is performed.
  • a related key word classification process 302 for integrating sets of single key word folders including similar document groups in regard to single key word folders prepared by the single key word classification process 301 is performed.
  • step 303 whether sets of related key word folders including similar document groups in regard to related key word folders prepared by the related key word classification process 302 can be integrated or not is judged.
  • the related key word folders are repeatedly integrated while the integration is possible. Whether detailed classification for information within the prepared single key word folders and related key word folders can be performed or not is examined (step 304) and hierarchical classification is repeated while detailed classification is possible (step 305).
  • the hierarchical classification is performed for information within all of the folders in step 305.
  • a folder prepared by the single key word classification process 301 is referred to as a single key word folder.
  • a table for managing key words is initialized.
  • the table for managing key words is referred to as a key word table.
  • a key word group of one document is read out from the key word data base 108 of Fig. 1 and each key word is registered in the key word table. Key words of all documents are repeatedly registered (step 403).
  • step 404 the number of documents having each key word included in the key word group is counted for each key word. Judgment as to whether the number of documents is one or not is made (step 405). When it is one, the key word is excluded from the object to be processed since the key word included in only one document is not useful for judgment of the relation to other documents in the classification.
  • a single key word folder having the same name as the key word is prepared and a document group having the key words included in the key word group is stored in the single key word folder. Further, the name of the single key word folder is registered in a table for managing the folder (step 407).
  • the table for managing the folder is referred to as a folder table.
  • step 408 the process for preparing the above single key word folder is repeatedly performed for all of the key words.
  • Each entry of the key word table 501 indicates a key word list 502.
  • the entry is determined by a value of a hash function having the key word as an input.
  • the key word list 502 includes a combination of a key word 503, a pointer 504 to the key word list having the same Hash value, and a pointer 505 to a document identification list.
  • the pointer 504 indicates a key word list 506 of the same Hash value and indicates that the key word of the key word list 506 having the same Hash value as the key word list 502 has the same Hash value.
  • the pointer 505 indicates a document identification number list 507 and is coupled with a document group including the key word 503.
  • the document identification number list 507 includes a combination of a number 508 for identifying a document and a pointer 509 to a next document identification number list.
  • the number of documents including key words is obtained by tracing the document identification number list 505 of the key word table 501.
  • a key word list is connected to a seventh entry 601 of a key word table 600.
  • a pointer 603 indicates that a key word list for a key word 609 of "new species" is connected to a key word list for a key word 602 of "pet” and a point 610 indicates that there is not a connection of a key word list any longer.
  • a pointer 604 indicates the document identification number 1 (605) which is one document identification number list including a key word of "pet” and a pointer 606 further indicates the document identification number 2 (607).
  • a pointer 608 indicates that there is not a document group including a key word of "pet” any longer.
  • a pointer 611 indicates the document identification number 3 (612) of the document identification number list including a key word of "new species”.
  • a pointer 613 indicates that the document identification number list is further connected.
  • the folder table 700 includes a combination of a folder name 701, the number of documents 702, a classification completion flag 703 and a document identification number list 704.
  • the classification completion flag 703 is a flag for storing whether the classification is performed or not and will be described in detail later.
  • the document identification number list 704 is equal to the document identification number list 507 of the key word table 501 of Fig. 5.
  • the 0-th entry 705 of the folder table 700 includes a folder of "pet", in which two documents are stored, and indicates that the classification of the folder is OFF (not completed) and the document identification number list is connected.
  • a document 801 is stored in the document data base 107 of Fig. 1 and forms a document group 802.
  • a key word group 803 is given to each document and is a gathering of individual key words 804.
  • Numeral 8031 denotes a key word group of a document 1 including key words of "dog”, “animal” and "pet”.
  • Numeral 8032 denotes a key word group of a document 2 including key words of "cat", “animal", "pet” and "mouse”.
  • Numeral 8033 denotes a key word group of a document 3 including key words of "new species”, “mouse”, “animal” and “ABC country”.
  • Numeral 8034 denotes a key word group of a document 4 including key words of "mouse”, “character goods” and "ABC country”.
  • Other documents 5, 6, 7 and 8 also include key word groups similarly.
  • the document group 802 is developed to a group of single key word folders 800.
  • the single key word folders are prepared with the same names as the key words of "dog”, “cat”, “character goods”, “new species”, “ABC country”, “pet”, “mouse” and “animal”, respectively.
  • Stored in each of the single key word folder is a collected document 805.
  • numeral 8051 denotes a collected document of a single key word folder "dog” including the document 1.
  • Numeral 8052 denotes a collected document of a single key word folder "pet” in which the documents 1 and 2 are stored.
  • Numeral 8053 denotes a collected document of a single key word folder "mouse” in which the documents 2, 3 and 4 are stored.
  • Numeral 8054 denotes a collected document of a single key word folder "animal” in which the documents 1, 2 and 3 are stored.
  • Numeral 8055 denotes a collected document of a single key word folder "ABC country” in which the documents 3 and 4 are stored.
  • Numeral 8056 denotes a collected document of a single key word folder "new species” in which the documents 3, 5, 6, 7 and 8 are stored.
  • a folder prepared by the related key word process 302 is hereinafter referred to as a related key word folder.
  • step 901 a table for managing the number of coincident documents between folders is initialized.
  • the table for managing the number of coincident documents between folders is referred to as an interfolder coincident document number management table.
  • Coincident documents between two single key word folders are counted and the count is registered in the interfolder coincident document number management table (step 902).
  • step 903 the number of coincident documents between all single key word folders are registered in the coincident document management table.
  • step 904 contents in the interfolder coincident document number management table are arranged in order of the descent of the number of coincident documents and whether a set of single key word folders at the top of the table, that is, a set of single key word folders having the maximum number of coincident documents between folders can be integrated or not is examined (step 905). Detail of step 905 is described later.
  • a related key word folder is prepared and document groups of two integrated single key word folders are stored therein.
  • a name of the related key word folder is registered in the folder table 700 of Fig. 7 and the two integrated single key word folders are deleted from the folder table 700 (step 907).
  • the name of the related key word folder is constituted by two combined names of the two integrated single key word folders.
  • step 908 completion of the integration is written in the interfolder coincident document number management table and the interfolder coincident document number management table is updated.
  • the integration process is repeated for single key word folders which are not subjected to the integration process from the top to the end of the interfolder coincident document number management table (step 906).
  • the interfolder coincident document number management table 1001 includes a combination of identification numbers 1002 and 1003 of two folders of which the number of coincident documents are examined, the number of coincident documents 1004 and an integrated flag 1005.
  • the number of coincident documents between the folder identification numbers 0 and 1 is five.
  • the integrated flag 1005 is a flag for storing "whether the integration of the folder is completed or not" or "the integration of the folder is impossible", and "integrated” means that the integration is completed and "impossible” means that the integration is impossible.
  • Fig. 11 shows a definite example of the interfolder coincident document number management table.
  • the identification numbers 1101 and the 1102 of all the single key word folders are listed and the number of coincident documents between two folders are investigated and registered. For example, it is shown that the number of coincident documents of documents 0 and 4 is 8.
  • the number of coincident documents is obtained by tracing and comparing the document identification list 704 of the folder table 700.
  • Fig. 12 shows a definite example of the update process of the inter folder coincident document number management table and description thereof is now made in detail.
  • the interfolder coincident document number management table 1200 shows a state that the processes have been completed until step 904 of Fig. 9.
  • the integration process of the folders is started from the top, that is, the 0-th entry of the interfolder coincident document number management table.
  • the integration is performed (step 905) and "integrated" is written in the integrated flag 1205 (step 908).
  • the integrated flags 1206 to 1213 relative to the numbers 0 and 1 are set to "impossible".
  • the integration process for the next entry or the 1st entry of the interfolder coincident document number management table is performed (step 906).
  • the integrated flag is not "impossible".
  • the integration process for one of the set of single key word folders of the 1st entry has been already performed and accordingly the integration cannot be performed.
  • the integrated flag 1214 is not "impossible"
  • the integration process is performed in order from the top of the interfolder coincident document number management table and accordingly the "integrated" entry does not appear behind the entry being processed.
  • a set of folders to be integrated is determined on condition that "the number of coincident documents is maximum". Whether the folders are to be integrated or not requires to examine whether the integrated result is an effective classification or not and judgment is made by comparing the state before integration with the integrated state. In the first embodiment, the distance calculation process of documents in a folder 1300 shown in Fig. 13 is applied.
  • a working key word table used temporarily in the distance calculation process 1300 of documents in a folder is initialized.
  • a data format of the working table is the same as the key word table 501 of Fig. 5.
  • Registration of key words is made for documents stored in the folder (step 1302) and is repeated for all documents in the folder (step 1303).
  • a key word number p initialized for counting of the number of key words (step 1304).
  • the key word having the number of documents equal to 1 is deleted from the working key word table (step 1308) and when the number of documents exceeds 1, the key word number p is incremented by one (step 1307).
  • the processes subsequent to the step 1305 are repeated (step 1309) and selection of key words included in two or more documents of key words included in the document group in the folder is completed. Then, the process proceeds to step 1400 of Fig. 14.
  • step 1401 of Fig. 14 a table for managing a word vector is initialized.
  • the importance degree is to be given by the word detection unit 102 of Fig. 1 and the larger the value the higher the importance degree.
  • the table for managing the word vector is referred to as a word vector table.
  • step 1402 a product of the appearance frequency of each key word and the importance degree of each key word given by the word detection unit 102 is calculated and registered in the word vector table. This process is repeated for each document (step 1403). Further, in step 1404, an average vector of word vectors in each document is calculated.
  • the distance between the vectors is to judge the nearness of the vectors and is defined in the first embodiment by the following equation.
  • Word vectors of documents Di and Dj are Wi and Wj, respectively, an angle between word vectors Wi and Wj is ⁇ and a distance between the documents Di and Dj is d(Di, Dj).
  • d(Di, Dj) 1-Wi ⁇ Wj/
  • 1-cos ⁇ where ⁇ represents the inner product, * the product and
  • the distance calculation of the word vector and the average vector is repeated for all the documents (step 1406).
  • a distance distribution is calculated from the distance of all average vectors and each document.
  • the distance distribution is defined by the following expression as (1) an average distance and (2) a variance.
  • the average value of the distance distribution investigated separately in respect to two single key word folders which are not integrated yet is compared with the distance distribution of the integrated related key word folder to thereby judge whether the integration is possible or not.
  • (1) and (2) are defined concretely by the following equation: (1)
  • represents an average value of x, and Td represents a threshold value) ⁇ 2 / ⁇ 1 > T ⁇ ( ⁇ 1 represents an average dispersion value of the two single key word folders which are not integrated yet, ⁇ 2 represents an integrated dispersion value, and T ⁇ represents a threshold value)
  • Td and T ⁇ are determined by an initial experiment and the values suitable for situation of the distance distribution of the document group in the folder are selected.
  • the word vector table 1500 is a two-dimensional table having the vertical axis indicating the document identification number and the horizontal axis indicating the key word identification number.
  • an entry 1503 indicates that a product of the appearance frequency and the importance degree of a key word of the key word identification number 3 in the document of the document identification number 3 is 2.
  • a calculation example of a distance of the word vector and the average vector is described with reference to Fig. 15. In order to simplify description, when the importance degree of each key word is all 1, the number of documents is 4, and the key words having the identification number 4 or less are processed, the distance of the word vector and the average vector for the document identification number 1 is calculated.
  • Fig. 16 shows a definite example of the related key word classification process.
  • a single key word folder "pet” 1601 and a single key word folder “animal” 1602 are integrated into a related key word folder "animal and pet” 1605, and a single key word folder "ABC country” 1603 and a single key word folder “mouse” 1604 are integrated into a related key word folder “mouse and ABC country” (step 302 of Fig. 3). Further, the related key word process is repeated (step 303 of Fig. 3) to integrate the related key word folder "animal and pet” 1605 and the related key word folder "mouse and ABC country” 1606 into a related key word folder "animal, mouse, pet and ABC country” 1607.
  • the name of the folder 1607 is described in order of "animal", “mouse", “pet” and “ABC country” to indicate that the number of documents related to the key words is larger in order of the description.
  • a first hierarchy or top class of the classification system is prepared.
  • the detailed classification process is now described.
  • the detailed classification is to classify the inside of the folder hierarchically and is made by using key words which are not utilized for preparation of the higher-rank folder.
  • step 1701 a type of the folder is judged.
  • the related key word folder is an integrated result of a plurality of single key word folders having high similarity, the inside of the folder is further classified hierarchically.
  • the classification recurrently repeats the single key word process 301, the related key word classification process 302, the integration process 303 of the related key word folders and the detailed classification 304 to 306 in accordance with the flow chart of Fig. 3.
  • a distance between documents in the folder is calculated as indicated in Fig. 13 (step 1300) and a variance value of the distance of each document and the average vector is calculated (step 1702).
  • the variance value is compared with the threshold T ⁇ (step 1702) and when the variance value is larger than the threshold T ⁇ , it is regarded that there is no merit to classify the inside of the folder hierarchically and the average distance is further examined (step 1703).
  • the document having the average distance exceeding the threshold Td is judged as a noise document having low relevance and is stored in a noise folder in the single key word (step 1704).
  • the noise folder is a folder for storing a document judged as noise and is prepared in only the folder in which a noise document exists.
  • the average distances for all the documents in the folder are calculated (step 1705) and the noise document is stored in a noise document folder.
  • Completion of classification is written in the classification completion flag 703 of the folder table 700 of Fig. 7 (step 1706).
  • a related key word folder "cat and fish” 1801 is classified into two folders of a single key word folder "cat food” 1804 and a related key word folder "fishing and sea” 1805 hierarchically. Further, a related key word folder "dog” 1802 is classified into tow folders of a single key word folder “midget Shiba” 1806 and a related key word folder "food and walk” 1807 hierarchically. In a single key word folder "new species” 1803, a noise document 1809 is divided into a noise document folder 1808.
  • the detailed classification has prepared the second and subsequent hierarchies of the classification system.
  • Each entry of the classification hierarchy management table 1900 indicates a folder information list.
  • the folder information list 1901 includes a set of a folder name 1902, a document identification number list 1903, the number of documents 1904, a pointer 1905 to a brother folder information list, a pointer 1906 to a child folder information list, and a pointer 1907 to a parent folder information list.
  • the folder name 1902 is coincident with the folder name 701 of the folder table 700 of Fig. 7 and the number of documents 1904 is coincident with the number of documents 702.
  • the document identification number list 1903 is a pointer to a document identification number list 1908 stored in each folder and the document identification number list 1908 includes a set of a document identification number 1909 and a pointer 1910 to a next document identification number list.
  • the document identification number list 1903 is coincident with the document identification number list 704 of the folder table 700 of Fig. 7.
  • the pointer 1905 to the brother folder information list has the same higher-rank folder as the folder information list 1901 and is a pointer to the folder information list positioned in the same hierarchy.
  • the pointer 1906 to the child folder information list is a point to one lower-rank information list of the folder information list 1901.
  • the pointer 1907 to the parent folder information list is a pointer to the higher-rank folder information list.
  • Contents of the folder table 700 is copied to the classification hierarchy management table 1900 at the same time when information of the folder prepared in the folder table 700 of Fig. 7 is written.
  • the pointer 1906 to the child folder information list, the pointer 1905 to the brother folder information list and the pointer 1907 to the parent folder information list of the classification hierarchy management table 1900 are updated.
  • a newly prepared folder is registered in the folder information list 1901. Further, when it is assumed that the inside of the folder is classified in detail into two folders, one child folder information list 1911 is registered in the pointer 1906 to the child folder information list and a second child folder information list 1915 is registered in a pointer 1912 to the brother folder information list of 1911. Since there is no further folder, 1916 indicates that there is no connection. Since the folder information list 1901 corresponds to the parent folder information list of the child folder information list 1911 and 1915, the pointers 1914 and 1918 to the parent folder information list indicate the folder information list 1901.
  • the folder information list 1901 Since the child folder information lists 1911 and 1915 of the folder information list 1901 have no hierarchical classification subsequent thereto, 1913 and 1918 indicate that there is no connection. Since the folder information list 1901 is a folder of a first hierarchy or top class and does not include any folder positioned in a further hierarchy or the same hierarchy, 1905 and 1907 indicate that there is no connection. In addition, the folder information list 1901 has two documents and 1903 and 1909 indicate first and second document identification number lists 1908 and 1910, respectively.
  • the classification system prepared by the document classification unit 103 of Fig. 1 is developed to a hierarchical structure as shown in Fig. 20 concretely.
  • the first hierarchy or top class there are the related key word folder "cat and fish” 1801, the single key word folder “dog” 1802 and the single key word folder “new species”.
  • the second hierarchical folder there are the single key word folder "cat food” 1804 and the related key word folder “fishing and sea” 1805 in a lower-rank of the folder 1801 and there are the single key word folder "midget Shiba" 1806 and the related key word folder "food and walk” 1807 in a lower rank of the folder 1802.
  • Noise document is separated into the noise document folder in a lower rank of 1803.
  • the classified result output unit 104 of Fig. 1 When the classified result output unit 104 of Fig. 1 receives the above classification system from the document classification unit 103, the classified result output unit prepares an interface screen as shown in Fig. 21 and outputs it to the CRT 109.
  • Three higher-rank hierarchies of the classification system are displayed in 2101 of Fig. 21, and 2113, 2114 and 2115 indicate the first hierarchy or top class, the second hierarchy and the third hierarchy, respectively.
  • the document group is displayed in the third hierarchy and a document 15 is selected.
  • 2112 indicates contents of the document 15, a prepared date 2116 of the document and the score information 2117 of the document 15 in the folder.
  • the user can select an interesting folder name by means of the mouse 111 of Fig. 1 and refer to contents thereof freely. Further, the user can skip uninteresting folders to reduce an amount of documents to refer to.
  • the document classification unit 103 can perform re-classification to cope with reception of new information.
  • the embodiment can assist to find out the key word.
  • the second embodiment concerns the document classification apparatus including classification designation means for constructing the classified result to which the user's opinion is reflected and provided in the classified result output unit 104 of the document classification apparatus 100 of the first embodiment shown in Fig. 1.
  • the document classification apparatus of the first embodiment sometimes produces the classified result different from the user's intention or idea since the document is classified automatically without restraint of the existing classification system. Accordingly, as shown in Fig. 22, a classification designation unit 2201 constituting means for reflecting the user's opinion to the classified result is added to the classified result output unit 104 of Fig. 1 to constitute a document classification apparatus 2200.
  • the classification designation unit 2201 includes two picture screens of a folder number designation interface 2500 and a classification system construction assisting interface 2700.
  • the document classification apparatus 2200 When the user specifies a document group desired to be classified by using the input device such as the key board 110 or the mouse 111 and directs to execute the classification process, the document classification apparatus 2200 is started to perform the process based on a flow chart of Fig. 23. First, in step 2301, the data management unit 101 stores the document group specified by the user into the document data base 107.
  • the word detection unit 102 detects key words from the document group and stores the key words into the word dictionary 105.
  • the document classification unit 103 creates the classification system in accordance with the flow chart of Fig. 3 in step 2303 and classifies the documents.
  • the classified result output unit 104 displays the classified result in the interface picture as shown in Fig. 21 to present it to the user (step 2304).
  • the processing steps described so far are the same as those of the first embodiment. Further, it is confirmed whether there is an indication to the classification designation unit 2201 from the user referring to the classified result or not (step 2305) and when there is no indication, the process is terminated.
  • step 2306 When there is an indication, contents of the indication are interpreted in step 2306, so that when it is the indication to the folder number designation interface 2500, a re-classification process of step 2307 is performed and when it is the indication to the classification system construct assisting interface 2700, a re-classification process of step 2308 is performed.
  • the folder number designation interface 2500 and the classification system construct assisting interface 2700 are described later in detail.
  • the process is returned to step 2304 and the re-classified result is presented to the user.
  • the case where the user issues an indication to the classification designation unit 2201 includes a case where the classified result does not satisfy the user's intention or idea.
  • Fig. 24 shows an interface picture indicating a classified result of a document group relative to computers of one thousand cases and prepared by the classified result output unit 104.
  • This interface picture is the same kind of output picture as Fig. 21 and the classified result prepared by the document classification unit 103 is read out and prepared in the same manner as the first embodiment.
  • information displayed in the picture of Fig. 14 is obtained by reading out the folder name 1902, the document group 1903, the number of documents 1904 and the hierarchical relation (1905, 1906 and 1907) of the folder from the classification hierarchy management table 1900 of Fig. 19.
  • Fig. 24 shows the three higher-rank hierarchies of the classification system by taking the case where several tens of folders are produced in the first hierarchy or top class as a result of classification and the equal number of folders are produced in the lower rank of each folder as an example.
  • a folder "personal computer, sale, selling and software" 2405 and a folder “sale, price, selling and prospects" 2406 are shown with the state thereof being reversed to indicate that the folders are selected in order that the user refers to the folders.
  • 2401 indicates vertically listed names of a plurality of folders produced in the first hierarchy or top class
  • 2402 indicates names of folders in the lower-rank (second) hierarchy of the user selected folder "personal computer, sale, selling and software" 2405 in the first hierarchy 2401
  • 2403 indicates vertically listed names of folders in the lower-rank (third) hierarchy of the user selected folder "sale, price, selling and prospects" 2406 of the second hierarchy 2402.
  • Fig. 2404 only six folder names are shown in the first hierarchy, while the picture can be scrolled by means of a scroll bar 2404 to refer to several folder names actually.
  • the key words “personal computer”, “sale”, “selling” and “software” contained in the folder “personal computer, sale, selling and software” 2405 of the first hierarchy are contained in the folder “sale, price, selling and prospects” 2406, the folder “personal computer, development, series and new product” 2407 and the folder “software, Tokyo, complexity and success” 2408 of the second hierarchy 2402.
  • the number of folders of the first hierarchy is increased in the classified result and it is difficult to find out a desired folder.
  • the number and the size of a folder proper as the classified result are different depending on the number of documents to be classified and the homogeneity of contents of the document. Since whether the classified result is proper or not is different in accordance with the viewpoint of the user who refers to the classified result, it is difficult to previously set the proper number of folders and the proper average number of documents.
  • the folder number designation interface 2500 serves to provide environment in which the number and the size of folders of the classified result produced by the document classification unit 103 can be designated by the user.
  • the folder number designation interface 2500 shown in Fig. 25 information relative to the classified result such as the number of folders and the average number of documents in the folder is presented to the user to receive an indication for integrating the number of folders and the average number of documents suitable in accordance with the user's desire again.
  • the number of folders and the average number of documents have the relation that both are interlocked with each other so that when the number of folders is reduced the average number of documents is increased, the user is adapted to be able to use either of the two as a reference for estimating the classified result.
  • 2501 represents the number of folders produced in one hierarchy and 2502 represents the average number of documents.
  • a folder number adjusting bar 2503 and an average document number adjusting bar 2504 are graphical user interfaces (GUI) for enabling the user to specify increase or reduction of the number of folders and the average number of documents.
  • 2505 represents the situation of the whole classification system by supposing that circles each having a diameter expressed in accordance with the number of documents are the folders.
  • the classification designation unit 2201 reads out information relative to the first hierarchy 2401 from the classified result output unit 104 and examines the number of folders and the number of documents in each folder to write the number of folders into 2501 and the average number of documents into 2502, so that the picture 2505 in which the folders are displayed by circles each having a radius proportional to the number of documents is prepared.
  • Fig. 25 shows a state that the folder number adjusting bar 2503 is operated by the user to indicate reduction of the number of folders.
  • the reclassification process 2307 is executed in response to the user's indication, a new classified result is produced and the folder number designation interface is changed from 2500 to 2510. Detail of the reclassification process 2307 is described later.
  • the number of folders 2501 is 96 before the re-classification, while the number of folders 2511 is reduced to 30 by reflecting the user's indication. In this manner, the user can adjust the classified result to a level that the user can easily refer to the number of folders.
  • Fig. 26 the re-classification process 2307 performed by the classification designation unit 2201 on the basis of the user's indication taken in from the folder number designation interface 2500 is described.
  • the classification designation unit 2201 analyzes the contents of the indication from the user in step 2601. When the indication is not to indicate "reduction of the number of folders" or "increase of the average number of documents", the classified result in the classification process 2303 is reviewed in step 2602 and the classification designation unit 2201 instructs the document classification unit 103 to re-select the classified result having the increased number of folders and the reduced average number of documents.
  • the re-selection process 2602 of the classified result performed by the document classification unit 103 is described.
  • the related key word process 302 is repeatedly applied on the basis of the flow chart of Fig. 3 to produce folders constituting the classified result and documents are classified into these folders.
  • the related key word process 302 is to integrate probably related sets of folders in accordance with the flow chart shown in Fig. 9 to thereby produce the folders constituting the classified result.
  • the folder table 700 (Fig. 7) for recording information relative to the folder is updated each time the related key word classification process 302 is repeated and information at the time that the related key word process 302 is completed (step 303) is adopted as the folder constituting the classified result.
  • the folders are integrated to reduce the number of folders and increase the average number of documents as a whole of the classified result. Accordingly, by recording information of the folder table 700 as intermediate folders for each of the related key word classification process 302, the folder having the increased number of folders and the reduced average number of documents as compared with the classified result can be re-selected later. For example, when it is assumed that the folder "animal, mouse, pet and ABC country" 1607 of Fig. 16 of the first embodiment is one of the folders constituting the classified result, the intermediate result includes the folder "animal and pet" 1605 and the folder "mouse and ABC country” 1606.
  • the intermediate result having "the smaller number of folders" and “the larger average number of documents” is selected in accordance with the increased and reduced level of the adjusting bar indicated by the user to thereby cope with the user's indication.
  • step 2601 when the user indicates "reduction of the number of folders” and "increase of the average number of documents", the classification designation unit 2201 instructs the document classification unit 103 to "remake the classified result having the smaller number of folders and the larger average number of documents" (step 2603).
  • the document classification unit 103 responds to this instruction to further integrate the folders of the classified result to thereby make "the classified result having the smaller number of folders and the larger average number of documents".
  • the case where the folders constituting the classified result are produced in the classification process 2303 is the case where it is judged in step 303 of the flow chart of Fig. 3 that the related key word classification process is completed, that is, the case where it is judged in step 905 of Fig.
  • the threshold values Td and T ⁇ constituting the integration judgment condition of step 905 are adjusted to set so that the folders can be integrated further.
  • the threshold Td is an index for judging that the integration is not proper when a distance from the average vector is separated over Td and accordingly, the threshold Td is reestablished to a larger value.
  • the threshold T ⁇ is an index for judging that the integration is not proper when a variance value of a distance from the average vector is larger than T ⁇ and accordingly the threshold T ⁇ is reestablished to a large value.
  • a definite value is determined by the document classification unit 103 in accordance with the increase and reduction level of the adjusting bar indicated by the user.
  • the related key word process 302 is applied to further integrate the folders repeatedly as much as possible (step 303). Consequently, since the classified result having the reduced number of folders and the increased average number of documents as compared with the classified result previously presented to the user can be prepared, the re-selection process 2602 of the classified result is performed from this result.
  • the classification system construction assisting interface 2700 which is another input picture presented by the classification designation unit 2201 is now described in detail.
  • the document classification unit 103 sometimes prepares the hierarchical relation which is inconsistent with the concept of general higher-rank words and lower-rank words since the document classification unit 103 hierarchizes the folders automatically without use of information such as the thesaurus.
  • the folder "dog" 1802 and the folder “midget Shiba” 1806 of Fig. 20 have the correct upper and lower relation, while if they are reversed, the reversed relation is inconsistent conceptually.
  • the classification system construction assisting interface 2700 receives auxiliary information for preparing a proper hierarchical structure from the user and allows the document classification unit 103 to perform the classification process by using the auxiliary information.
  • auxiliary information for preparing a proper hierarchical structure from the user and allows the document classification unit 103 to perform the classification process by using the auxiliary information.
  • particular items such as "Title of the Invention” and “Claims” and contents to be described in each item are defined and these information corresponds to the auxiliary information.
  • the contents of description in each item are different in each specification, while it is considered that there are common points in the degree of the specialty and generality and the degree is determined for each item.
  • key words appearing in items of contents having the higher generality are utilized to prepare folders of the higher-rank hierarchy and key words appearing in items of contents having the higher specialty are utilized to prepare folders of the lower-rank hierarchy to thereby be apt to construct the proper hierarchical structure.
  • Items utilized in the classification are limited for each hierarchy so that key words appearing in the items such as "Title of the Invention” and “Field of the Invention” having the higher generality are utilized to prepare the higher-rank hierarchy and key words appearing in the items such as "Summary of the Invention” and "Description of the Preferred Embodiments” are utilized to prepare the lower-rank hierarchy.
  • the classification system construction assisting interface 2700 can easily receive the items included in the document and the specialty of the description in the items from the user.
  • Fig. 27 shows an displayed example of the classification system construction assisting interface 2700 in the case where the document classification apparatus 2200 classifies a specification of a patent.
  • the process of receiving the auxiliary information from the user through the classification system construction assisting interface 2700 by the classification designation unit 2201 is now described with reference to a flow chart of Fig. 28.
  • the classification designation unit 2201 reads out a sample document to set it to a picture 2701.
  • the sample document is one of documents stored in the document data base 107 and is a specification of a patent in Fig. 27 by way of example.
  • information relative to items in the sample document is received from the user.
  • the user can specify a string of characters on the picture 2701 by means of the mouse 2702.
  • the character string of the item "Title of the Invention” is dragged by means of the mouse 2702 to be specified as an item.
  • a dialogue box 2706 is displayed in step 2803 to require the user to confirm whether the item specified by the user is proper or not.
  • information relative to the item and an appearance position of the item on the sample document is taken in (step 2805).
  • the classification designation unit 2201 stores the item of "Title of the Invention" and the information of the appearance position of "the first to the fifth locations of characters”.
  • the items are arranged in order of appearance thereof in the sample document on the basis of the appearance positions of the items in step 2806 and the arranged items are displayed in 2712 of an item setting picture 2710 (step 2807).
  • information relative to the degree of the specialty of all the items is received from the user.
  • the classification designation unit 2201 displays the previously prepared degrees of several kinds indicative to the specialty in a level list 2714. Accordingly, the user selects one of the items from 2712 by using the displayed level list 2714 and selects one proper level from the level list 2714 in consideration of the generality and the specialty of the contents thereof. Then, the selection is settled by means of an OK button 2715. The designation is performed by such a procedure.
  • the item setting picture 2710 provides the five kinds of degrees of the specialty relative to the items such as "general”, “rather general”, “toss-up”, “rather special” and “special” and accordingly the degree suitable for each item is selected therefrom.
  • the degree of the specialty relative to the items is referred to as an item level, which will be described in detail later.
  • the classification designation unit 2201 prepares hierarchical construction information on the basis of the item and the item level taken in from the user.
  • the hierarchical construction information is information which the classification process 2303 refers to in order that the document classification unit 103 constructs the classified result of a proper hierarchical structure and is to prescribe the items and the importance of the items to be utilized in the classification when a certain hierarchy is constructed. Detail thereof is described later.
  • next step 2902 the hierarchical construction information is used to prepare a temporary key word data base.
  • the temporary key word data base is to store a set of key words to which the importance is given as the key word utilized in the classification when a certain hierarchy is constructed and has the same format as the key word data base of the first embodiment. Preparation of the temporary key word is described later.
  • the above process is a processing step for preparing data for constructing the classified result of a proper hierarchical structure on the basis of the information taken in by the classification designation unit 2201 from the user.
  • the classification designation unit 2201 instructs the document classification unit 103 to construct the classified result of the proper hierarchical structure by using the above data.
  • the document classification unit 103 uses the temporary key word data base in place of the key word data base of the first embodiment to perform the single key word process 301 for clustering documents having key words included in the key word group for each key word.
  • step 2904 the related key word process 302 for integrating sets of single key word folders including similar document groups is performed in regard to single key word folders prepared by the single key word classification process 2903.
  • the hierarchical construction information is used to give the importance to each key word of the word vector. Detail thereof is described later.
  • next step 303 in the same manner as the first embodiment, the integration of the related key word folders including similar document groups is repeated for the related key word folders prepared by the related key word process. Further, whether the prepared single key word folder and related key word folder can be classified in detail or not is examined (in the same manner as in step 304 of the first embodiment) and when the detailed classification is possible, the inside of the folder is classified (step 2905). In step 2905, the process B from step 2902 to the end of Fig. 29 is applied to the documents classified in the folder. In step 306, the hierarchical classification is repeated for the inside of all the folders.
  • the document classification apparatus 2200 is assumed to prepare the classification system of three hierarchies and five kinds of item levels including "general”, “rather general”, “toss-up”, “rather special” and “special” can be set.
  • Four items of Fig. 30 and the item levels are given by the user through the classification system construction assisting interface 2700 in accordance with the flow chart of Fig. 28.
  • rules for preparing the hierarchical construction information the following are supposed, for example.
  • the hierarchical construction information for the item level of Fig. 30 is prepared on the basis of the above rules as follows:
  • step 2902 in the case of constructing the first hierarchy or top class, the first hierarchical construction information is used to prepare the temporary key word data base as follows:
  • the key word group of documents are read out from the key word data base 108 and only key words appearing in the items "first chapter” and "fourth chapter” are taken out to prepare the temporary key word group and register it in the temporary key word data base.
  • a key word group of a document 3100 of Fig. 31 is registered as a key word group 3110 in the key word data base 108.
  • these taken-out key words correspond to key words a 3105, b 3106, g 3107 and h 3108.
  • the temporary key word group 3112 is prepared from these key words and is registered in the temporary key word data base 3111.
  • the single key word classification 2903 utilizes the temporary key word group thus prepared.
  • the related key word classification process 2904 performed using the hierarchical construction information and the temporary key word data base is now described.
  • the related key word process 2904 for constructing the first hierarchy top class
  • the fact that the key words appearing in the item "first chapter” have the importance degree a and the key words appearing in the item "fourth chapter” have the importance degree b is read from the first hierarchical construction information.
  • the above importance degrees are then added up to the appearance frequencies of the key words included in each document to increase the importance degrees of the key words to be attached importance upon the first hierarchical construction.
  • the key words appearing in the items specified by the user are attached importance and classified to thereby be apt to prepare the folders suitable for the hierarchy, so that the classification system having the proper higher and lower relation is prepared.
  • a third embodiment of a document classification apparatus 4100 (shown in Fig. 41) which previously receives instructions relative to the classification from the user in the classification designation unit 2201 described in the second embodiment and performs the classification process on the basis of the received instructions is now described.
  • the document classification apparatus 4100 When the user specifies a document group desired to be classified by using the input device such as the key board 110 or the mouse 111 and instructs execution of the classification process, the document classification apparatus 4100 is started to perform the process based on a flow chart of Fig. 32. In step 2301, the data management unit 101 stores the user's specified document group into the document data base 107.
  • next step 2302 the word detection unit 102 detects key words from the document group and stores them into the word dictionary 105.
  • the process waits for the user's instructions.
  • the user's instructions are delivered to the document classification unit 103 to perform the classification process 3201.
  • step 2304 the classified result output unit 104 displays the classified result. Further, the classification designation unit 2201 confirms the instructions from the user who refers to the classified result (step 2305) and when there are not any instructions, the process is terminated. When there are the user's instructions, contents of the instructions are interpreted in step 2306 and when the they are instructions to the folder number designation interface 2500, the re-classification process of step 2307 is performed while when they are instructions to the classification system construction assisting interface 2500, the re-classification process of step 2308 is applied. As an example where there are the user's instructions relative to the classification again in step 2305, there is considered the case where the classified result previously specified by the user does not suit the user's convenience contrary to the user's expectations.
  • Fig. 33 is a flow chart showing a definite process of the classification process 3201.
  • step 3301 it is examined whether instructions are issued to the folder number designation interface 2500 or to the classification system construction assisting interface 2700.
  • the re-classification process 2308 of the classification system construction assisting interface 2700 is performed and the process is terminated.
  • the document classification unit 103 prepares the classification system in accordance with the flow chart of Fig. 3 and classifies documents in step 2303.
  • step 3302 it is confirmed whether the classified result satisfies the user instructed conditions or not.
  • the process is terminated.
  • the reclassification process 2307 of the folder number designation interface 2500 is executed and the classification process 2303 is executed again to satisfy the user's instructions.
  • a fourth embodiment of a document classification apparatus including means for causing the user to browse the classified result and provided in the classified result output unit 104 of the document classification apparatus 100 of the first embodiment shown in Fig. 1 is now described.
  • the folders prepared as the classified result are quite unknown differently from the case where classification is made on the basis of the previously set classification system. Accordingly, when the user refers to the classified result to retrieve a desired document, it is examined what kind of folder there is. Thus, the user selects a folder in which there is a possibility that a desired document is classified therein and browses the inside of the folder. However, if it takes time to reach the desired document, the user easily browses the inside of the folder when the classification system is fixed. With such a system, the present invention which prepares a proper classification system in accordance with the document to be classified is not utilized effectively. Accordingly, as shown in Fig.
  • a document classification apparatus 3400 including a retrieval supporting unit 3401 in addition to the document classification apparatus of the first embodiment is provided to support browsing of the classified result.
  • the retrieval supporting unit 3401 includes two processing functions of a folder retrieval supporting function 3402 and a document retrieval supporting function 3403.
  • the document classification apparatus 3400 When the user specifies a document group desired to be classified by using the input device such as the key board 110 or the mouse 111 and instructs to execute the classification process, the document classification apparatus 3400 is started to perform the process based on a flow chart of Fig. 35. In step 2301, the data management unit 101 stores the document group specified by the user into the document data base 107.
  • the word detection unit 102 detects key words from the document group and stores them into the word dictionary 105.
  • the document classification unit 103 prepares the classification system in accordance with the flow chart of Fig. 3 and classifies documents.
  • the classified result output unit 104 displays the classified result in an interface picture 3600 as shown in Fig. 36 to present it to the user (step 2304).
  • a folder name of the first hierarchy or top class is displayed in 3601
  • a folder name of the second hierarchy is displayed in 3602
  • a folder name of the third hierarchy is displayed in 3603.
  • the processing steps described so far are the same as those of the first and second embodiments.
  • step 3501 it is examine whether the user selects any folder or document of the classification situations displayed by the classified result output unit 104 or not.
  • the process is terminated (step 3502).
  • the folder retrieval supporting process 3504 is indicated to the folder retrieval supporting function 3402 of the retrieval supporting unit 3401 and when the document is selected, the document retrieval supporting process 3505 is indicated to the document retrieval supporting function 3403 of the retrieval supporting unit 3401.
  • the folder retrieval supporting process 3504 and the document retrieval supporting process 3505 are to process the classified result so that the classified result is apt to be browsed and detail thereof is described later. Further, when the user presses a search supporting button 3605 to require to support browsing of the classified result, the retrieval supporting unit 3401 delivers the processed classified result to the classified result output unit 104 to display the new classified result.
  • the folder retrieval supporting function 3402 is to estimate other folders on the basis of the similarity to the folder selected by the user in order that the user refers thereto and re-arrange the folders in order of the similarity.
  • Fig. 36 shows the interface picture 3600 of the classified result produced by the classified result output unit 104 and an auxiliary information picture 3610 produced by the retrieval supporting unit 3401.
  • the folder names of the first hierarchy or top class, the second hierarchy and the third hierarchy are displayed in the interface picture 3600 and the user can select and refer to them.
  • the folder name is abstracted and shown as "folder a", while the folder name is actually constituted by one or more key words.
  • Auxiliary information relative to the classified result such as the number of folders 3611 and the average number of documents 3612 is displayed in the auxiliary information picture 3610.
  • the folders prepared in the first hierarchy are displayed in the picture 3613 with circles having diameters corresponding to the number of documents thereof and are arranged in order of the number of documents.
  • the user selects a desired folder on the basis of the information and refers to the documents classified in the folder.
  • the classified result output unit 104 informs the folder retrieval supporting function 3402 of the retrieval supporting unit 3401 of it to execute the folder retrieval supporting process 3504.
  • the folder retrieval supporting process 3504 the similarity of the folder a and other folders of the first hierarchy is examined. The similarity between the folders is judged so that "the more the documents in a folder coincident with the documents classified in the folder a, the higher the similarity of the folder". This is obtained in the related key word process 302 of the document classification unit 103 described in the first embodiment. That is, in the fourth embodiment, this corresponds to the classification process of step 3501 of Fig. 35.
  • the folder retrieval supporting process 3504 since the similarity between the folders is required, the information relative to the number of coincident documents between the folders is stored in the classification process 3501 of the fourth embodiment and is provided to the folder retrieval supporting function 3402. Concretely, in step 902 of Fig. 9, contents of the interfolder coincident document number management table 1001 are stored.
  • the interfolder coincident document number management table 1001 in order to examine the similarity of the folder a and other folders, reference is first made to the interfolder coincident document number management table 1001 to take out only a column of the identification number corresponding to the folder a from the folder identification numbers 1002 and 1003. Since the interfolder coincident document number management table 1001 is arranged in order of descent on the basis of the coincident document number, the number of documents coincident with the folder a is increased in a column positioned at the higher-rank location of the table in the taken-out column and the similarity thereof is high. That is, information relative to the folder of the first hierarchy can be taken out in order of the similarity to the folder a.
  • the folder retrieval supporting function 3402 uses these information and rewrites the picture 3613 of the auxiliary information picture 3610 into a picture 3701 of an auxiliary information picture 3700 of Fig. 37. That is, the folders of the first hierarchy are rearranged in order of similarity to the folder a so that the user easily obtains information relative to the folder similar to the folder a.
  • the document retrieval supporting function 3403 is now described.
  • the document retrieval supporting function 3403 is to examine a document similar to a document judged as improper by the user of the documents classified in the folder and exclude the document from the folder to thereby support the retrieval operation of a desired document.
  • Fig. 38 shows an interface picture 3800 of the classified result produced by the classified result output unit 104 and an auxiliary information picture 3810 produced by the retrieval supporting unit 3401.
  • folder names are displayed in the first hierarchy or top class 3801
  • folder names of the second hierarchy relative to the folder a selected by the user are displayed in 3802
  • a list of document names relative to a sub-folder a is displayed in 3803.
  • the user selects a document a from 3803 and refers to the document a in a picture 3804 in which contents of the document are displayed.
  • a list of documents in the sub-folder a including the document a referred currently by the user and top portions of contents of the documents are displayed in 3811 of the auxiliary information picture 3810.
  • the user browses the classified result on the basis of these information.
  • the retrieval supporting unit 3401 informs the document retrieval supporting function 3403 of this information.
  • the document retrieval supporting function 3403 performs the document retrieval supporting process 3505 in accordance with a flow chart of Fig. 39.
  • the distance calculation process 1405 between word vectors described in the first embodiment is used.
  • the distance calculation process of documents in a folder 1300 is performed for the sub-folder a.
  • step 1401 of Fig. 14 the word vector table 1500 is initialized and preparation of word vectors in each document is repeated (step 1402 and 1403).
  • the processing steps described so far are the same process as the first embodiment.
  • the distance calculation process 1405 between word vectors of the first embodiment is performed for the document a and other documents. A distance between documents is similarity and accordingly the similarity between the document a and each document is calculated.
  • documents having a distance between the documents and the document a shorter than the threshold Ta, that is, document having the close similarity to the document a are examined and these documents are regarded as improper document candidates (step 3903).
  • the above process is repeated for the whole documents (step 3904) to select all the documents having the high similarity to the document a.
  • the retrieval supporting unit 3401 delivers the information relative to the document judged as the improper document candidates to the classified result output unit 104. Thereafter, when the user specifies the retrieval supporting button 3805, the document similar to the document a is excluded from the folder and an interface picture 400 of the classified result of Fig. 40 is prepared newly.
  • the document a and the documents judged as similar to the document a are removed from the sub-folder a and the number of documents is changed from 96 cases (3806) of Fig. 38 to 71 cases (4001) of Fig. 40. 25 documents including the document a and documents d, f and the like judged as similar to the document a are removed from the list of documents 3803 and the list of documents is changed as 4002.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
EP95115253A 1994-09-30 1995-09-27 Procédé et dispositif pour classer des informations de documents Expired - Lifetime EP0704810B1 (fr)

Applications Claiming Priority (6)

Application Number Priority Date Filing Date Title
JP23644494 1994-09-30
JP23644494 1994-09-30
JP236444/94 1994-09-30
JP23103395 1995-09-08
JP23103395A JP3669016B2 (ja) 1994-09-30 1995-09-08 文書情報分類装置
JP231033/95 1995-09-08

Publications (2)

Publication Number Publication Date
EP0704810A1 true EP0704810A1 (fr) 1996-04-03
EP0704810B1 EP0704810B1 (fr) 2002-04-03

Family

ID=26529664

Family Applications (1)

Application Number Title Priority Date Filing Date
EP95115253A Expired - Lifetime EP0704810B1 (fr) 1994-09-30 1995-09-27 Procédé et dispositif pour classer des informations de documents

Country Status (4)

Country Link
US (1) US5832470A (fr)
EP (1) EP0704810B1 (fr)
JP (1) JP3669016B2 (fr)
DE (1) DE69526168T2 (fr)

Cited By (21)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1997038382A1 (fr) * 1996-04-03 1997-10-16 Siemens Aktiengesellschaft Procede de classification automatique d'un texte reporte sur un document apres avoir ete transforme en donnees numeriques
WO1997041519A1 (fr) * 1996-04-29 1997-11-06 Gosudarstvenny Nauchno-Tekhnichesky Tsentr Giperinformatsionnykh Tekhnology (Gntts 'gintekh') Procedes de traitement automatique de materiaux porteurs d'informations en vue d'une utilisation personnalisee
EP0820027A2 (fr) * 1996-07-18 1998-01-21 Matsushita Electric Industrial Co., Ltd. Procédé de support et appareil de recouvrement de données et médium de stockage pour le programme support de recouvrement de données
EP0822503A1 (fr) * 1996-08-02 1998-02-04 Matsushita Electric Industrial Co., Ltd. Système de recouvrement de documents
US6098066A (en) * 1997-06-13 2000-08-01 Sun Microsystems, Inc. Method and apparatus for searching for documents stored within a document directory hierarchy
US6185550B1 (en) * 1997-06-13 2001-02-06 Sun Microsystems, Inc. Method and apparatus for classifying documents within a class hierarchy creating term vector, term file and relevance ranking
US6199071B1 (en) 1997-04-01 2001-03-06 Sun Microsystems, Inc. Method and apparatus for archiving hypertext documents
US6219665B1 (en) 1997-06-06 2001-04-17 Matsushita Electric Industrial Co., Ltd. Retrieval menu creation device, a retrieval menu creation method, and a recording medium that stores a retrieval menu creation program
WO2002025479A1 (fr) * 2000-09-25 2002-03-28 Telstra New Wave Pty Ltd Systeme de categorisation de documents
WO2002031697A1 (fr) * 2000-10-13 2002-04-18 Kent Ridge Digital Labs Procédé pour visualiser des bouquets de grandes collections de documents textuels
EP1304627A2 (fr) 2001-10-19 2003-04-23 Xerox Corporation Procédés, systèmes et dispositifs pour un regroupement hiérarchique doux d'objets réoccurants
WO2003052627A1 (fr) * 2001-12-18 2003-06-26 Telstra New Wave Pty Ltd Taxinomie de ressources d'informations
EP1324218A1 (fr) * 2001-12-11 2003-07-02 Abb Research Ltd. Système de categoriser des objets de donnés et procédé de vérifier la consistance des categories designees aux objets d'information
US6990496B1 (en) 2000-07-26 2006-01-24 Koninklijke Philips Electronics N.V. System and method for automated classification of text by time slicing
CN101976246A (zh) * 2010-09-30 2011-02-16 互动在线（北京）科技有限公司 百科词条分类检索方法
CN102902674A (zh) * 2011-07-26 2013-01-30 腾讯科技（深圳）有限公司 服务群组分类方法和系统
EP2061034B1 (fr) * 2003-06-30 2013-10-09 FUJIFILM Corporation Procédé de gestion de fichiers et dispositif d'imagerie
CN103377216A (zh) * 2012-04-24 2013-10-30 苏州引角信息科技有限公司 产品信息库的构建方法及系统
US8601397B2 (en) 2005-03-31 2013-12-03 Barclays Capital Inc. System and method for grouping a collection of documents using document series
CN104123279A (zh) * 2013-04-24 2014-10-29 腾讯科技（深圳）有限公司 关键词的聚类方法和装置
CN104765726A (zh) * 2015-04-27 2015-07-08 湘潭大学 一种基于信息密度的数据分类方法

Families Citing this family (151)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH11505342A (ja) * 1995-02-27 1999-05-18 アンフォルマシオン メディカル エ スタティスティーク 情報オブジェクト、特に文書、を処理しアクセスする方法および該方法を実施するためのシステム
US6069706A (en) 1995-07-31 2000-05-30 Canon Kabushiki Kaisha Image reading device and image processing method utilizing the same
US5867164A (en) * 1995-09-29 1999-02-02 Apple Computer, Inc. Interactive document summarization
JPH10143517A (ja) * 1996-11-13 1998-05-29 Fuji Xerox Co Ltd 文書作成装置
JPH10228486A (ja) * 1997-02-14 1998-08-25 Nec Corp 分散ドキュメント分類システム及びプログラムを記録した機械読み取り可能な記録媒体
JP3173411B2 (ja) * 1997-03-17 2001-06-04 富士ゼロックス株式会社 関連文書検索装置及び関連文書検索プログラムを記録した記録媒体
EP0972254A1 (fr) * 1997-04-01 2000-01-19 Yeong Kuang Oon Procede de traitement de texte didactique et oriente contenu comportant un systeme de croyances modifie de fa on incrementielle
JP3001460B2 (ja) 1997-05-21 2000-01-24 株式会社エヌイーシー情報システムズ 文書分類装置
US6820094B1 (en) * 1997-10-08 2004-11-16 Scansoft, Inc. Computer-based document management system
JPH11134359A (ja) * 1997-10-31 1999-05-21 Nippon Telegr & Teleph Corp <Ntt> 文書類似度計算方法、文書類似度計算装置及び文書類似度計算プログラムを記録した記録媒体
US6542888B2 (en) * 1997-11-26 2003-04-01 International Business Machines Corporation Content filtering for electronic documents generated in multiple foreign languages
US6014662A (en) * 1997-11-26 2000-01-11 International Business Machines Corporation Configurable briefing presentations of search results on a graphical interface
US6389436B1 (en) * 1997-12-15 2002-05-14 International Business Machines Corporation Enhanced hypertext categorization using hyperlinks
JP3571515B2 (ja) * 1997-12-19 2004-09-29 富士通株式会社 知識収集・格納・検索プログラムを記憶したコンピュータ読み取り可能な記憶媒体
WO1999042936A1 (fr) * 1998-02-24 1999-08-26 Gateway 2000, Inc. Systeme de gestion de logiciels
JPH11242679A (ja) * 1998-02-25 1999-09-07 Nippon Telegr & Teleph Corp <Ntt> 利用者の興味に基づいて情報を分類する方法、装置、および利用者の興味に基づいて情報を分類するプログラムを記録した記録媒体
JPH11259504A (ja) * 1998-03-11 1999-09-24 Mitsubishi Electric Corp データベース作成装置およびデータベース検索装置
US6269188B1 (en) 1998-03-12 2001-07-31 Canon Kabushiki Kaisha Word grouping accuracy value generation
JPH11288417A (ja) * 1998-04-01 1999-10-19 Fujitsu Ltd ドキュメント管理システム
US6415282B1 (en) 1998-04-22 2002-07-02 Nec Usa, Inc. Method and apparatus for query refinement
US6243501B1 (en) 1998-05-20 2001-06-05 Canon Kabushiki Kaisha Adaptive recognition of documents using layout attributes
JP3622503B2 (ja) 1998-05-29 2005-02-23 株式会社日立製作所 特徴文字列抽出方法および装置とこれを用いた類似文書検索方法および装置並びに特徴文字列抽出プログラムを格納した記憶媒体および類似文書検索プログラムを格納した記憶媒体
JP3665480B2 (ja) 1998-06-24 2005-06-29 富士通株式会社 文書整理装置および方法
US7152031B1 (en) 2000-02-25 2006-12-19 Novell, Inc. Construction, manipulation, and comparison of a multi-dimensional semantic space
US7197451B1 (en) 1998-07-02 2007-03-27 Novell, Inc. Method and mechanism for the creation, maintenance, and comparison of semantic abstracts
US7039856B2 (en) * 1998-09-30 2006-05-02 Ricoh Co., Ltd. Automatic document classification using text and images
JP2000285140A (ja) * 1998-12-24 2000-10-13 Ricoh Co Ltd 文書処理装置、文書分類装置、文書処理方法、文書分類方法およびそれらの方法をコンピュータに実行させるプログラムを記録したコンピュータ読み取り可能な記録媒体
US6349307B1 (en) * 1998-12-28 2002-02-19 U.S. Philips Corporation Cooperative topical servers with automatic prefiltering and routing
US6853950B1 (en) * 1999-07-20 2005-02-08 Newsedge Corporation System for determining changes in the relative interest of subjects
US6397215B1 (en) 1999-10-29 2002-05-28 International Business Machines Corporation Method and system for automatic comparison of text classifications
US6754660B1 (en) * 1999-11-30 2004-06-22 International Business Machines Corp. Arrangement of information for display into a continuum ranging from closely related to distantly related to a reference piece of information
SG108244A1 (en) * 2000-04-13 2005-01-28 Ibm A method and apparatus for search and display of hierarchically organized information of arbitrary depth using graphical user interfaces
US6408277B1 (en) 2000-06-21 2002-06-18 Banter Limited System and method for automatic task prioritization
US9699129B1 (en) 2000-06-21 2017-07-04 International Business Machines Corporation System and method for increasing email productivity
US8290768B1 (en) * 2000-06-21 2012-10-16 International Business Machines Corporation System and method for determining a set of attributes based on content of communications
US6925608B1 (en) * 2000-07-05 2005-08-02 Kendyl A. Roman Graphical user interface for building Boolean queries and viewing search results
US7286977B1 (en) * 2000-09-05 2007-10-23 Novell, Inc. Intentional-stance characterization of a general content stream or repository
US7672952B2 (en) * 2000-07-13 2010-03-02 Novell, Inc. System and method of semantic correlation of rich content
US7389225B1 (en) 2000-10-18 2008-06-17 Novell, Inc. Method and mechanism for superpositioning state vectors in a semantic abstract
US7653530B2 (en) * 2000-07-13 2010-01-26 Novell, Inc. Method and mechanism for the creation, maintenance, and comparison of semantic abstracts
US6778972B2 (en) 2000-08-10 2004-08-17 Gustavo S. Leonardos′ System and method for providing integrated management of electronic information
US7603415B1 (en) * 2000-08-15 2009-10-13 ART Technology Group Classification of electronic messages using a hierarchy of rule sets
US7177922B1 (en) 2000-09-05 2007-02-13 Novell, Inc. Policy enforcement using the semantic characterization of traffic
US20090234718A1 (en) * 2000-09-05 2009-09-17 Novell, Inc. Predictive service systems using emotion detection
US20100122312A1 (en) * 2008-11-07 2010-05-13 Novell, Inc. Predictive service systems
US6560600B1 (en) * 2000-10-25 2003-05-06 Alta Vista Company Method and apparatus for ranking Web page search results
DE10057634C2 (de) * 2000-11-21 2003-01-30 Bosch Gmbh Robert Verfahren zur Verarbeitung von Text in einer Rechnereinheit und Rechnereinheit
JP3696498B2 (ja) * 2000-11-24 2005-09-21 パイオニア株式会社 情報編集装置
US7644057B2 (en) * 2001-01-03 2010-01-05 International Business Machines Corporation System and method for electronic communication management
US6938206B2 (en) * 2001-01-19 2005-08-30 Transolutions, Inc. System and method for creating a clinical resume
JP2002222083A (ja) 2001-01-29 2002-08-09 Fujitsu Ltd 事例蓄積装置および方法
US6826576B2 (en) 2001-05-07 2004-11-30 Microsoft Corporation Very-large-scale automatic categorizer for web content
US7536413B1 (en) 2001-05-07 2009-05-19 Ixreveal, Inc. Concept-based categorization of unstructured objects
USRE46973E1 (en) 2001-05-07 2018-07-31 Ureveal, Inc. Method, system, and computer program product for concept-based multi-dimensional analysis of unstructured information
US7194483B1 (en) 2001-05-07 2007-03-20 Intelligenxia, Inc. Method, system, and computer program product for concept-based multi-dimensional analysis of unstructured information
US6970881B1 (en) 2001-05-07 2005-11-29 Intelligenxia, Inc. Concept-based method and system for dynamically analyzing unstructured information
US7627588B1 (en) 2001-05-07 2009-12-01 Ixreveal, Inc. System and method for concept based analysis of unstructured data
JP2003016109A (ja) * 2001-06-29 2003-01-17 Hitachi Ltd 文書情報管理方法および装置、および管理サーバ
US7010515B2 (en) * 2001-07-12 2006-03-07 Matsushita Electric Industrial Co., Ltd. Text comparison apparatus
JP4261783B2 (ja) * 2001-09-11 2009-04-30 キヤノン株式会社 文書登録システム、方法、プログラム及び記憶媒体
KR20030024296A (ko) * 2001-09-17 2003-03-26 (주)넷피아닷컴 리얼 네임을 이용한 웹 페이지 접속 시스템 및 그 방법
KR20030024297A (ko) * 2001-09-17 2003-03-26 (주)넷피아닷컴 검색 시스템 및 그 방법
US20030061132A1 (en) * 2001-09-26 2003-03-27 Yu, Mason K. System and method for categorizing, aggregating and analyzing payment transactions data
US7305402B2 (en) * 2001-10-10 2007-12-04 International Business Machines Corporation Adaptive indexing technique for use with electronic objects
US20030069892A1 (en) * 2001-10-10 2003-04-10 International Business Machines Corporation Relational view of electronic objects
US6985908B2 (en) * 2001-11-01 2006-01-10 Matsushita Electric Industrial Co., Ltd. Text classification apparatus
JP2003150602A (ja) * 2001-11-15 2003-05-23 Hitachi Ltd 文書情報管理方法および装置
US20030154181A1 (en) * 2002-01-25 2003-08-14 Nec Usa, Inc. Document clustering with cluster refinement and model selection capabilities
US8589413B1 (en) 2002-03-01 2013-11-19 Ixreveal, Inc. Concept-based method and system for dynamically analyzing results from search engines
JP4082059B2 (ja) * 2002-03-29 2008-04-30 ソニー株式会社 情報処理装置および方法、記録媒体、並びにプログラム
JP2004110445A (ja) * 2002-09-19 2004-04-08 Hitachi Ltd 文書管理方法、プログラムおよびシステム
JP4514401B2 (ja) * 2002-10-10 2010-07-28 株式会社ジャストシステム 文書分類装置、文書分類方法およびその方法をコンピュータに実行させるプログラム
US20040088157A1 (en) * 2002-10-30 2004-05-06 Motorola, Inc. Method for characterizing/classifying a document
US7158962B2 (en) 2002-11-27 2007-01-02 International Business Machines Corporation System and method for automatically linking items with multiple attributes to multiple levels of folders within a content management system
US8473532B1 (en) * 2003-08-12 2013-06-25 Louisiana Tech University Research Foundation Method and apparatus for automatic organization for computer files
US7769759B1 (en) * 2003-08-28 2010-08-03 Biz360, Inc. Data classification based on point-of-view dependency
TWI290687B (en) * 2003-09-19 2007-12-01 Hon Hai Prec Ind Co Ltd System and method for search information based on classifications of synonymous words
US7516492B1 (en) * 2003-10-28 2009-04-07 Rsa Security Inc. Inferring document and content sensitivity from public account accessibility
JP4239090B2 (ja) 2004-01-08 2009-03-18 富士フイルム株式会社 ファイル管理プログラム
CA2558404A1 (fr) 2004-03-02 2005-09-15 Volt Information Sciences Inc. Procede et un systeme pour le transfert d'information d'affaires de revendeur consultant
JP4461854B2 (ja) * 2004-03-16 2010-05-12 富士ゼロックス株式会社 文書分類装置ならびに文書分類装置の分類体系生成装置および方法
US8775436B1 (en) * 2004-03-19 2014-07-08 Google Inc. Image selection for news search
US20050246333A1 (en) * 2004-04-30 2005-11-03 Jiang-Liang Hou Method and apparatus for classifying documents
US7370273B2 (en) * 2004-06-30 2008-05-06 International Business Machines Corporation System and method for creating dynamic folder hierarchies
JP3716272B2 (ja) * 2004-07-02 2005-11-16 パイオニア株式会社 情報編集装置及び情報編集方法
US7580921B2 (en) * 2004-07-26 2009-08-25 Google Inc. Phrase identification in an information retrieval system
US7711679B2 (en) * 2004-07-26 2010-05-04 Google Inc. Phrase-based detection of duplicate documents in an information retrieval system
US7702618B1 (en) 2004-07-26 2010-04-20 Google Inc. Information retrieval system for archiving multiple document versions
US7567959B2 (en) 2004-07-26 2009-07-28 Google Inc. Multiple index based information retrieval system
JP2006065366A (ja) * 2004-08-24 2006-03-09 Nec Corp キーワード分類装置およびその方法、端末装置ならびにプログラム
TWI254880B (en) * 2004-10-18 2006-05-11 Avectec Com Inc Method for classifying electronic document analysis
US20060203258A1 (en) * 2005-03-10 2006-09-14 Kabushiki Kaisha Toshiba File management apparatus
US20060206531A1 (en) * 2005-03-10 2006-09-14 Kabushiki Kaisha Toshiba Document managing apparatus
JP2006301959A (ja) * 2005-04-20 2006-11-02 Just Syst Corp 文書処理装置、文書処理方法、文書処理プログラムおよびコンピュータに読み取り可能な記録媒体
US8112401B2 (en) * 2005-05-25 2012-02-07 Ecteon, Inc. Analyzing externally generated documents in document management system
US8447781B2 (en) * 2005-07-29 2013-05-21 International Business Machines Corporation Content-based file system security
EP1952280B8 (fr) 2005-10-11 2016-11-30 Ureveal, Inc. Systeme, procede et produit-programme d'ordinateur pour recherche et analyse conceptuelles
US9069847B2 (en) * 2005-10-21 2015-06-30 Battelle Memorial Institute Data visualization methods, data visualization devices, data visualization apparatuses, and articles of manufacture
WO2007050646A2 (fr) * 2005-10-24 2007-05-03 Capsilon Fsg, Inc. Procede d'entreprises pour le traitement automatise de papiers et de documents electroniques non structures
US8176004B2 (en) * 2005-10-24 2012-05-08 Capsilon Corporation Systems and methods for intelligent paperless document management
US7676485B2 (en) 2006-01-20 2010-03-09 Ixreveal, Inc. Method and computer program product for converting ontologies into concept semantic networks
US20070185860A1 (en) * 2006-01-24 2007-08-09 Michael Lissack System for searching
KR100756921B1 (ko) * 2006-02-28 2007-09-07 한국과학기술원 문서 분류방법 및 그 문서 분류방법을 컴퓨터에서 실행시키기 위한 프로그램을 포함하는 컴퓨터로 읽을 수있는 기록매체.
GB0612433D0 (en) * 2006-06-23 2006-08-02 Ibm Method and system for defining a hierarchical structure
US7809723B2 (en) * 2006-06-26 2010-10-05 Microsoft Corporation Distributed hierarchical text classification framework
US7873583B2 (en) * 2007-01-19 2011-01-18 Microsoft Corporation Combining resilient classifiers
JP4967705B2 (ja) * 2007-02-22 2012-07-04 富士ゼロックス株式会社 クラスタ生成装置およびクラスタ生成プログラム
US8117223B2 (en) * 2007-09-07 2012-02-14 Google Inc. Integrating external related phrase information into a phrase-based indexing information retrieval system
JP4524702B2 (ja) 2008-02-08 2010-08-18 コニカミノルタビジネステクノロジーズ株式会社 データ管理装置、検索条件情報管理方法、およびコンピュータプログラム
JP4626662B2 (ja) 2008-03-21 2011-02-09 ブラザー工業株式会社 データ保存装置及びコンピュータプログラム
US8661069B1 (en) * 2008-03-31 2014-02-25 Google Inc. Predictive-based clustering with representative redirect targets
CN101661461B (zh) 2008-08-29 2016-01-13 阿里巴巴集团控股有限公司 确定文档中核心地理信息的方法、系统
US8296297B2 (en) * 2008-12-30 2012-10-23 Novell, Inc. Content analysis and correlation
US8301622B2 (en) * 2008-12-30 2012-10-30 Novell, Inc. Identity analysis and correlation
US8386475B2 (en) * 2008-12-30 2013-02-26 Novell, Inc. Attribution analysis and correlation
JP5322660B2 (ja) * 2009-01-07 2013-10-23 キヤノン株式会社 データ表示装置、データ表示方法、コンピュータプログラム
JP5564816B2 (ja) * 2009-03-31 2014-08-06 日本電気株式会社 階層型文書管理システム、階層型文書管理方法および階層型文書管理プログラム
US20100250479A1 (en) * 2009-03-31 2010-09-30 Novell, Inc. Intellectual property discovery and mapping systems and methods
US9245243B2 (en) 2009-04-14 2016-01-26 Ureveal, Inc. Concept-based analysis of structured and unstructured data using concept inheritance
JP4620165B2 (ja) * 2009-08-10 2011-01-26 富士フイルム株式会社 ファイル管理システム
US8788357B2 (en) 2009-08-12 2014-07-22 Iqnavigator, Inc. System and method for productizing human capital labor employment positions/jobs
JP5434500B2 (ja) * 2009-11-13 2014-03-05 富士ゼロックス株式会社 木構造処理装置及びプログラム
US20110295861A1 (en) * 2010-05-26 2011-12-01 Cpa Global Patent Research Limited Searching using taxonomy
JP5126289B2 (ja) * 2010-06-03 2013-01-23 コニカミノルタビジネステクノロジーズ株式会社 検索フォルダの管理方法、フォルダの管理方法、コンピュータ、およびコンピュータプログラム
US9679256B2 (en) * 2010-10-06 2017-06-13 The Chancellor, Masters And Scholars Of The University Of Cambridge Automated assessment of examination scripts
US8396871B2 (en) 2011-01-26 2013-03-12 DiscoverReady LLC Document classification and characterization
JP5751086B2 (ja) * 2011-08-12 2015-07-22 日本電気株式会社 コンテンツ管理装置およびコンテンツ管理方法
US10467252B1 (en) * 2012-01-30 2019-11-05 DiscoverReady LLC Document classification and characterization using human judgment, tiered similarity analysis and language/concept analysis
US9667514B1 (en) 2012-01-30 2017-05-30 DiscoverReady LLC Electronic discovery system with statistical sampling
US9659087B2 (en) * 2012-11-19 2017-05-23 Amplero, Inc. Unsupervised prioritization and visualization of clusters
JP6132617B2 (ja) * 2013-03-26 2017-05-24 キヤノン株式会社 受信した画像データをフォルダに格納する画像処理システム、画像処理方法、及びプログラム
US9275331B2 (en) * 2013-05-22 2016-03-01 International Business Machines Corporation Document classification system with user-defined rules
US11336648B2 (en) 2013-11-11 2022-05-17 Amazon Technologies, Inc. Document management and collaboration system
US10599753B1 (en) 2013-11-11 2020-03-24 Amazon Technologies, Inc. Document version control in collaborative environment
US10540404B1 (en) * 2014-02-07 2020-01-21 Amazon Technologies, Inc. Forming a document collection in a document management and collaboration system
US9542391B1 (en) 2013-11-11 2017-01-10 Amazon Technologies, Inc. Processing service requests for non-transactional databases
US10691877B1 (en) 2014-02-07 2020-06-23 Amazon Technologies, Inc. Homogenous insertion of interactions into documents
US10635645B1 (en) 2014-05-04 2020-04-28 Veritas Technologies Llc Systems and methods for maintaining aggregate tables in databases
KR20160015838A (ko) * 2014-07-31 2016-02-15 삼성전자주식회사 컨텐츠 분류 방법 및 이를 위한 디바이스
US9807073B1 (en) 2014-09-29 2017-10-31 Amazon Technologies, Inc. Access to documents in a document management and collaboration system
US10095768B2 (en) * 2014-11-14 2018-10-09 Veritas Technologies Llc Systems and methods for aggregating information-asset classifications
CN104731890B (zh) * 2015-03-16 2017-12-12 山东师范大学 一种结合plsa和at的音频事件分类方法
JP6540268B2 (ja) * 2015-06-24 2019-07-10 富士ゼロックス株式会社 オブジェクト分類装置及びプログラム
US10055489B2 (en) * 2016-02-08 2018-08-21 Ebay Inc. System and method for content-based media analysis
JP6963372B2 (ja) * 2016-10-18 2021-11-10 株式会社東芝 名刺情報管理システムおよび検索結果表示プログラム
JP6930179B2 (ja) * 2017-03-30 2021-09-01 富士通株式会社 学習装置、学習方法及び学習プログラム
CN108563747A (zh) * 2018-04-13 2018-09-21 北京深度智耀科技有限公司 一种文档处理方法及装置
WO2020204535A1 (fr) * 2019-03-29 2020-10-08 주식회사 워트인텔리전스 Procédé, dispositif et système de classification automatique personnalisée par utilisateur de documents de brevet sur la base d'un apprentissage machine
US11687717B2 (en) * 2019-12-03 2023-06-27 Morgan State University System and method for monitoring and routing of computer traffic for cyber threat risk embedded in electronic documents
KR102277643B1 (ko) 2021-02-05 2021-07-14 한국기술교육대학교 산학협력단 계층적 다중 레이블 분류를 위한 데이터 분류 방법 및 장치
JP2023023590A (ja) * 2021-08-05 2023-02-16 京セラドキュメントソリューションズ株式会社 画像処理装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0437615A1 (fr) * 1989-06-14 1991-07-24 Hitachi, Ltd. Procede de recherche documentaire a prerecherche hierarchique, appareil a cet effet, et dispositif a disque magnetique destine a cet appareil
EP0457707A2 (fr) * 1990-05-16 1991-11-21 International Business Machines Corporation Procédé pour grouper hiérarchiquement les documents dans un système de traitement de données
JPH0528198A (ja) 1991-07-22 1993-02-05 Hiroyuki Sone 文書情報検索装置
EP0542429A2 (fr) * 1991-11-12 1993-05-19 Xerox Corporation Procédé et dispositif basé sur groupes pour regarder de grandes collections de documents
JPH05324726A (ja) 1992-05-25 1993-12-07 Fujitsu Ltd 文書データ分類装置及び文書分類機能構築装置

Family Cites Families (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4719571A (en) * 1986-03-05 1988-01-12 International Business Machines Corporation Algorithm for constructing tree structured classifiers
US4760606A (en) * 1986-06-30 1988-07-26 Wang Laboratories, Inc. Digital imaging file processing system
US4958284A (en) * 1988-12-06 1990-09-18 Npd Group, Inc. Open ended question analysis system and method
US5508912A (en) * 1989-01-23 1996-04-16 Barry Schneiderman Clinical database of classified out-patients for tracking primary care outcome
JP2993673B2 (ja) * 1989-01-27 1999-12-20 株式会社日立製作所 電子ファイル装置
US5201047A (en) * 1989-12-21 1993-04-06 International Business Machines Corporation Attribute-based classification and retrieval system
JPH0490054A (ja) * 1990-08-03 1992-03-24 Toshiba Corp 画像記憶検索装置
US5590317A (en) * 1992-05-27 1996-12-31 Hitachi, Ltd. Document information compression and retrieval system and document information registration and retrieval method
US5355497A (en) * 1992-06-10 1994-10-11 Physiotronics Corporation File directory structure generator and retrevial tool with document locator module mapping the directory structure of files to a real world hierarchical file structure
JP3210102B2 (ja) * 1992-11-17 2001-09-17 松下電器産業株式会社 電子化文書ファイリング装置及びファイリング文書検索方法
JPH0744568A (ja) * 1993-07-30 1995-02-14 Mitsubishi Electric Corp 検索インタフェース装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0437615A1 (fr) * 1989-06-14 1991-07-24 Hitachi, Ltd. Procede de recherche documentaire a prerecherche hierarchique, appareil a cet effet, et dispositif a disque magnetique destine a cet appareil
EP0457707A2 (fr) * 1990-05-16 1991-11-21 International Business Machines Corporation Procédé pour grouper hiérarchiquement les documents dans un système de traitement de données
JPH0528198A (ja) 1991-07-22 1993-02-05 Hiroyuki Sone 文書情報検索装置
EP0542429A2 (fr) * 1991-11-12 1993-05-19 Xerox Corporation Procédé et dispositif basé sur groupes pour regarder de grandes collections de documents
JPH05324726A (ja) 1992-05-25 1993-12-07 Fujitsu Ltd 文書データ分類装置及び文書分類機能構築装置

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
"Information Society in the 21st Century", November 1991, NIKKEI BYTE, pages: 320 - 331
GROUP WARE RESEARCH MEETING, vol. 4-6, 28 October 1993 (1993-10-28), pages 37 - 44
INFORMATION PROCESSING SOCIETY OF JAPAN, article "Broad Catch System of Information", pages: 13 - 6

Cited By (36)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6295543B1 (en) * 1996-04-03 2001-09-25 Siemens Aktiengesellshaft Method of automatically classifying a text appearing in a document when said text has been converted into digital data
WO1997038382A1 (fr) * 1996-04-03 1997-10-16 Siemens Aktiengesellschaft Procede de classification automatique d'un texte reporte sur un document apres avoir ete transforme en donnees numeriques
WO1997041519A1 (fr) * 1996-04-29 1997-11-06 Gosudarstvenny Nauchno-Tekhnichesky Tsentr Giperinformatsionnykh Tekhnology (Gntts 'gintekh') Procedes de traitement automatique de materiaux porteurs d'informations en vue d'une utilisation personnalisee
US6009439A (en) * 1996-07-18 1999-12-28 Matsushita Electric Industrial Co., Ltd. Data retrieval support apparatus, data retrieval support method and medium storing data retrieval support program
EP0820027A2 (fr) * 1996-07-18 1998-01-21 Matsushita Electric Industrial Co., Ltd. Procédé de support et appareil de recouvrement de données et médium de stockage pour le programme support de recouvrement de données
EP0820027A3 (fr) * 1996-07-18 1998-11-04 Matsushita Electric Industrial Co., Ltd. Procédé de support et appareil de recouvrement de données et médium de stockage pour le programme support de recouvrement de données
EP0822503A1 (fr) * 1996-08-02 1998-02-04 Matsushita Electric Industrial Co., Ltd. Système de recouvrement de documents
US6199071B1 (en) 1997-04-01 2001-03-06 Sun Microsystems, Inc. Method and apparatus for archiving hypertext documents
US6219665B1 (en) 1997-06-06 2001-04-17 Matsushita Electric Industrial Co., Ltd. Retrieval menu creation device, a retrieval menu creation method, and a recording medium that stores a retrieval menu creation program
US6098066A (en) * 1997-06-13 2000-08-01 Sun Microsystems, Inc. Method and apparatus for searching for documents stored within a document directory hierarchy
US6185550B1 (en) * 1997-06-13 2001-02-06 Sun Microsystems, Inc. Method and apparatus for classifying documents within a class hierarchy creating term vector, term file and relevance ranking
US6990496B1 (en) 2000-07-26 2006-01-24 Koninklijke Philips Electronics N.V. System and method for automated classification of text by time slicing
WO2002025479A1 (fr) * 2000-09-25 2002-03-28 Telstra New Wave Pty Ltd Systeme de categorisation de documents
EP1323078A4 (fr) * 2000-09-25 2005-03-02 Telstra New Wave Pty Ltd Systeme de categorisation de documents
US7971150B2 (en) 2000-09-25 2011-06-28 Telstra New Wave Pty Ltd. Document categorisation system
EP1323078A1 (fr) * 2000-09-25 2003-07-02 Telstra New Wave Pty Ltd Systeme de categorisation de documents
WO2002031697A1 (fr) * 2000-10-13 2002-04-18 Kent Ridge Digital Labs Procédé pour visualiser des bouquets de grandes collections de documents textuels
US7644102B2 (en) 2001-10-19 2010-01-05 Xerox Corporation Methods, systems, and articles of manufacture for soft hierarchical clustering of co-occurring objects
EP1304627A2 (fr) 2001-10-19 2003-04-23 Xerox Corporation Procédés, systèmes et dispositifs pour un regroupement hiérarchique doux d'objets réoccurants
EP1304627A3 (fr) * 2001-10-19 2007-03-07 Xerox Corporation Procédés, systèmes et dispositifs pour un regroupement hiérarchique doux d'objets réoccurants
WO2003054727A1 (fr) * 2001-12-11 2003-07-03 Abb Research Ltd Systeme de categorisation pour des objets de donnees et procede pour verifier la coherence d'affectations d'objets de donnees a des categories
EP1324218A1 (fr) * 2001-12-11 2003-07-02 Abb Research Ltd. Système de categoriser des objets de donnés et procédé de vérifier la consistance des categories designees aux objets d'information
EP1456774A1 (fr) * 2001-12-18 2004-09-15 Telstra New Wave Pty Ltd Taxinomie de ressources d'informations
EP1456774A4 (fr) * 2001-12-18 2005-02-23 Telstra New Wave Pty Ltd Taxinomie de ressources d'informations
WO2003052627A1 (fr) * 2001-12-18 2003-06-26 Telstra New Wave Pty Ltd Taxinomie de ressources d'informations
AU2002347222B2 (en) * 2001-12-18 2008-05-29 Telstra Corporation Limited Information resource taxonomy
EP2061034B1 (fr) * 2003-06-30 2013-10-09 FUJIFILM Corporation Procédé de gestion de fichiers et dispositif d'imagerie
US8601397B2 (en) 2005-03-31 2013-12-03 Barclays Capital Inc. System and method for grouping a collection of documents using document series
CN101976246A (zh) * 2010-09-30 2011-02-16 互动在线（北京）科技有限公司 百科词条分类检索方法
CN102902674B (zh) * 2011-07-26 2017-09-26 腾讯科技（深圳）有限公司 服务群组分类方法和系统
CN102902674A (zh) * 2011-07-26 2013-01-30 腾讯科技（深圳）有限公司 服务群组分类方法和系统
CN103377216A (zh) * 2012-04-24 2013-10-30 苏州引角信息科技有限公司 产品信息库的构建方法及系统
CN104123279A (zh) * 2013-04-24 2014-10-29 腾讯科技（深圳）有限公司 关键词的聚类方法和装置
CN104123279B (zh) * 2013-04-24 2018-12-07 腾讯科技（深圳）有限公司 关键词的聚类方法和装置
CN104765726A (zh) * 2015-04-27 2015-07-08 湘潭大学 一种基于信息密度的数据分类方法
CN104765726B (zh) * 2015-04-27 2018-07-31 湘潭大学 一种基于信息密度的数据分类方法

Also Published As

Publication number Publication date
EP0704810B1 (fr) 2002-04-03
JPH08153121A (ja) 1996-06-11
DE69526168T2 (de) 2002-11-28
JP3669016B2 (ja) 2005-07-06
DE69526168D1 (de) 2002-05-08
US5832470A (en) 1998-11-03

Similar Documents

Publication Publication Date Title
US5832470A (en) Method and apparatus for classifying document information
JP3577819B2 (ja) 情報探索装置及び情報探索方法
Marcos-Pablos et al. Information retrieval methodology for aiding scientific database search
JP3942290B2 (ja) ドキュメント画像のクライアント・ワークステーションへの送信方法
JP4335335B2 (ja) ドキュメント画像のソート方法
JP3303926B2 (ja) 構造化文書分類装置及び方法
US6327586B1 (en) System method and computer program product to automate the management and analysis of heterogeneous data
US7194471B1 (en) Document classification system and method for classifying a document according to contents of the document
US7331517B2 (en) Article reader program, article management method and article reader
EP0722145B1 (fr) Système de recouvrement de données et procédé d&#39;opération
US5950214A (en) System, method, and computer program product for accessing a note database having subnote information for the purpose of manipulating subnotes linked to portions of documents
US5799325A (en) System, method, and computer program product for generating equivalent text files
US6772148B2 (en) Classification of information sources using graphic structures
JP3614618B2 (ja) 文献検索支援方法及び装置およびこれを用いた文献検索サービス
US7734634B2 (en) System, apparatus and method for using and managing digital information
EP0731948B1 (fr) Procede et appareil pour synchroniser, afficher et manipuler des documents contenant du texte et des images
JPH11224345A (ja) ドキュメント画像の一部の識別方法
US20090070101A1 (en) Device for automatically creating information analysis report, program for automatically creating information analysis report, and method for automatically creating information analysis report
US20060179051A1 (en) Methods and apparatus for steering the analyses of collections of documents
GB2350712A (en) Document processor and recording medium
US20130326338A1 (en) Methods and systems for organizing content using tags and for laying out images
WO2000054185A1 (fr) Procede et dispositif d&#39;elaboration d&#39;un thesaurus a l&#39;usage d&#39;utilisateurs, au moyen de bases de donnees en ligne
US20050114302A1 (en) Method for fast searching and displaying a genealogical tree of patents from a patent database
JP2004348771A (ja) 技術文書検索装置
Barros et al. Indexing multispectral images for content-based retrieval

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

AK Designated contracting states

Kind code of ref document: A1

Designated state(s): DE FR GB

17P Request for examination filed

Effective date: 19960919

17Q First examination report despatched

Effective date: 19991007

GRAG Despatch of communication of intention to grant

Free format text: ORIGINAL CODE: EPIDOS AGRA

GRAG Despatch of communication of intention to grant

Free format text: ORIGINAL CODE: EPIDOS AGRA

GRAH Despatch of communication of intention to grant a patent

Free format text: ORIGINAL CODE: EPIDOS IGRA

REG Reference to a national code

Ref country code: GB

Ref legal event code: IF02

GRAH Despatch of communication of intention to grant a patent

Free format text: ORIGINAL CODE: EPIDOS IGRA

GRAA (expected) grant

Free format text: ORIGINAL CODE: 0009210

AK Designated contracting states

Kind code of ref document: B1

Designated state(s): DE FR GB

REF Corresponds to:

Ref document number: 69526168

Country of ref document: DE

Date of ref document: 20020508

ET Fr: translation filed
PLBE No opposition filed within time limit

Free format text: ORIGINAL CODE: 0009261

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: NO OPPOSITION FILED WITHIN TIME LIMIT

26N No opposition filed

Effective date: 20030106

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: GB

Payment date: 20030826

Year of fee payment: 9

Ref country code: FR

Payment date: 20030826

Year of fee payment: 9

PGFP Annual fee paid to national office [announced via postgrant information from national office to epo]

Ref country code: DE

Payment date: 20030908

Year of fee payment: 9

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: GB

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20040927

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: DE

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20050401

GBPC Gb: european patent ceased through non-payment of renewal fee

Effective date: 20040927

PG25 Lapsed in a contracting state [announced via postgrant information from national office to epo]

Ref country code: FR

Free format text: LAPSE BECAUSE OF NON-PAYMENT OF DUE FEES

Effective date: 20050531

REG Reference to a national code

Ref country code: FR

Ref legal event code: ST